Kernel

What do they actually do

Kernel runs Chrome/Chromium browsers for you in the cloud. Developers spin up a browser via an API/SDK or the dashboard, receive a Chrome DevTools (CDP) websocket URL, and connect existing tools like Playwright or Puppeteer to drive that remote browser. Each session can stream a live view, produce video replays, and emit logs for observing and debugging automations in real time [docs].

It supports reusing cookies and full session state so automations stay logged in across hours or days. Teams can optionally host their agent code on Kernel and invoke it as an endpoint instead of running their own infra. The underlying browser images are open source so customers can inspect or run them locally [docs] [kernel-images].

Kernel advertises very fast cold starts (roughly sub-300–325 ms) and a usage-based model that charges by active browser time, with plan-based limits for concurrency and persistence [homepage] [pricing].

Who are their target customer(s)

• AI‑agent startups and teams building web‑facing automation: They need browsers that start quickly, don’t flake under load, and are observable so agents can interact with real sites without spending engineering time building and operating a browser fleet.
• QA and end‑to‑end testing teams: They want reproducible runs and easy debugging (video replays, live view, logs) because slow, brittle environments and manual triage delay releases.
• Back‑office automation / RPA engineers: They need long‑lived sessions and reliable auth persistence so bots can stay logged in across hours or days without brittle cookie handling or frequent reauthentication.
• Data‑collection and web‑monitoring companies that require real browsers: They need predictable concurrency, proxy/IP controls, and the ability to inspect or run the exact browser images to avoid rendering mismatches and scale reliably.
• Regulated enterprise teams (security/compliance owners): They prefer a managed option to reduce compliance and operational burden, but require enterprise controls, BYO options, and formal compliance guarantees before production use.

How would they acquire their first 10, 50, and 100 customers

• First 10: Target YC/accelerator and GitHub communities already using Playwright/Puppeteer; offer credits and hands‑on onboarding to integrate their agents so Kernel becomes the easiest path to production.
• First 50: Publish turnkey templates and SDK tutorials, ship short demo videos, and list marketplace/integration entries (e.g., Vercel) to meet developers where they work; run targeted outreach to QA/RPA teams with paid trials and fast support.
• First 100: Add a small SDR/AE motion to convert higher‑ACV accounts, emphasize enterprise controls (BYO images, proxies, compliance), use case studies and customer referrals, and form channel partnerships with AI and observability platforms.

What is the rough total addressable market

Top-down context:

Kernel sells into adjacent markets that commonly need real browsers: RPA (~$18.2B in 2024), automation testing (~$28B in 2023, growing), and web data/extraction (~$0.7B). Combined, that implies roughly ~$47B of relevant spend today [Fortune Business Insights] [MarketsandMarkets coverage] [ScrapingDog].

Bottom-up calculation:

If ~30,000 QA/RPA/data/agent teams worldwide require managed real browsers and 20% adopt a hosted service at an average $25k/year, the near‑term SAM is ~${150}M; at 40% adoption the SAM approaches ~${300}M. Capturing a few percent of the broader ~$47B pool would imply a low‑billion revenue ceiling over time.

Assumptions:

• A meaningful subset of QA/RPA/web‑data teams prefer hosted browsers to self‑managed fleets for reliability, speed, and compliance.
• Average customer spend of ~$25k/year across a mix of SMB, mid‑market, and early enterprise buyers.
• Legal/anti‑bot constraints limit some scraping use cases, modestly reducing the serviceable market.

Who are some of their notable competitors

• Browserless: Direct browsers‑as‑a‑service provider that plugs into Puppeteer/Playwright; closest in shape to Kernel with hosted Chrome instances, session features, and live debugging for automation teams.
• Apify: Cloud platform for web scraping/automation that runs user “actors,” with ready‑made scrapers, storage, proxies, and scheduling—competes when customers want full scraping pipelines, not just a low‑latency browser runtime.
• BrowserStack: Large cross‑browser testing cloud with broad device/OS coverage and enterprise test workflows; overlaps when teams prioritize comprehensive test orchestration over agent‑focused, low‑latency sessions.
• Sauce Labs: Enterprise test and automation platform offering virtual device/cloud testing, recording, and debugging at scale; strong choice for compliance and long‑standing enterprise SLAs.
• LambdaTest: Cloud automation grid for Selenium/Playwright/Cypress with parallel execution and CI/CD integrations; aimed at QA scale testing more than long‑running agent sessions or hosted agent runtimes.